This paper was originally published in Privredna kretanja i ekonomska politika (Economic Trends and*

Economic Policy), No. 95, 2003, pp. 27-46.

An earlier version of this paper was presented at the European Regional Science Association Conference "Peripheries,Centres and Spatial Development in the New Europe" held on August 27-30, 2003 in Jyväskylä, Finland. Theauthor wish to thank the Conference participants, and in particular Professor Jouke van Dijk for their usefulcomments. Author's thanks also go to Andrea Mervar and Sandra Švaljek for their helpful advice during thedevelopment of a draft version of this paper, as well as to Danijel Nestiæ for his constructive suggestions. The authortakes responsibility for any remaining errors or omissions.

Valerija Botriæ, Institute of Economics, Zagreb.**

77

4 Regional Aspects

of Unemployment

in Croatia*

Abstract

This paper analyzes regional labor market differences in Croatia. The unemployment rates

are identified for five Croatian regions, which are formed according to the NUTS II

introduction proposal. Differences between regional and national unemployment dynamics

are investigated by applying seemingly unrelated regression methods. The results indicate

that Croatian regions can be divided into three groups, with Central Croatia having the

strongest correlation with developments at the national level.

Keywords: regional unemployment, regression analysis, seemingly unrelated regression

equations, Croatia.

JEL classification: R23, J64

See, for example, Bertola, Boeri and Nicoletti (2001, p. 210).1

Amendola, Caroleo and Coppola (2003).2

78

Introduction

The issue of relatively high unemployment rates has frequently been addressed in the

analysis of the Croatian economy. Nonetheless, numerous questions relating to the

labor market have not been sufficiently explored empirically. The issue of scarce

empirical analysis is even more pronounced at a regional level.

What makes the study of regional labor market differences so interesting is the context

of the European economic integration. According to theory, a country decides to enter

a monetary union after having met the following conditions:

a) external influences on total demand and supply in countries participating in

the union should be symmetrical,

b) the mobility of the labor force should be higher within the member countries,

c) real wages should be flexible.1

Conditions for a successful integration are obviously related to the labor market. One

of the guiding principles of the overall integration process declares that the

introduction of a common market will contribute to an increased mobility not only

of capital, to which the state borders have already ceased to present an obstacle, but also

of the labor force. An increased mobility will contribute to a more efficient allocation

of resources through market competition. However, the answer to the question whether

such theoretical predictions can be proven in practice is far from unequivocal. Research

indicates that there are regions in some countries which have, over longer periods,

achieved inferior results in the labor market relative to the national average. If the

mobility of production factors is not ensured at the national level, the mobility cannot

be expected to increase simply thanks to the influence of wider integration processes.

Thus actual benefits of the integration may be far less significant than expected.

Empirical research indicates that differences among EU regions are much more

profound than among individual member states. Therefore, an interesting question2

arises as to whether further divergence between the regions may occur as a consequence

of the integration process since some regions, as opposed to others, may benefit more

Gacs and Huber (2003).3

79

from the process itself. In this context regional differences in Croatia need to be

compared with the movements in other EU candidate countries. Current empirical

research indicates that the differences between regional labor markets of candidate

countries increase during the transition period and approximate those of Western

Europe. The question is whether such differences will keep growing or whether the3

trend will reverse. Long-lasting divergent movements (conditional or unconditional) are

incompatible with theoretical models, particularly with neoclassical growth models.

However, labor market models, while predicting a gradual disappearance of regional

differences, monitor the restoring of balance speed. According to the latter models, the

rate at which the labor market restores its balance is commensurate to the labor market

flexibility; flexibility is assessed through labor force mobility and the possibility of wage

adjustments to market conditions. The requirement for increased flexibility in the labor

market, often quoted in public, arises directly from these theoretical models.

The experience of other transition countries, in particular of the EU candidate

countries, is also relevant to Croatia. However, a comparison of the Croatian labor

market relative to other countries cannot be carried out without identifying certain

specifics or peculiarities. One of these peculiarities that affect the labor market in

Croatia is the recent war experience. The impact of the war itself is a sufficient reason

to test whether the results of other studies could be deemed relevant in Croatian case.

The following section presents a theoretical model, which addresses the issue of

regional differences in the labor market and specifies the segment of the theoretical

framework discussed in the paper. The third section discusses the regional data

availability issue in Croatia, and its implications with regard to the interpretation of the

results. The results of empirical research and their interpretation are outlined in the

fourth section while the fifth and last section is a conclusion.

Theoretical Framework

Regional aspects of the labor market are frequently observed in the literature within a

model articulated by Blanchard and Katz (1992). The model incorporates

developments regarding employment, unemployment, activity and wage rates in

tRtRtRwzl

,,,

α−=

tRtRitRwz

,,1,0,ζρρ +−=∆

tRtRtRppopn

,,,

+=

tRtRtRtRwup ,,2,10, ζλλλ +++=

The short presentation of the model has also been used by Gacs and Huber (2003).4

80

individual regions, and interprets their interrelationship in a given region. The basic

idea behind the models is presented next. According to this model, employment

movement in a specific region may be illustrated by the following equation:4

(1)

where l stands for employment in the region R over the time period t ; w stands forR,t R,t

wages in the region R over the time period t ; while z is the parameter explaining shifts

in the labor demand curve. In other words, employment depends upon a specific

position on the labor demand curve (the z parameter primarily involves capital

mobility) as well as upon wages. All the model variables have been expressed in

logarithmic differences from the national average, because the model aims to explain

why some regions exhibit better indicators relative to the national average, and to

project movements of such differences in time.

The decision of an entrepreneur on the location of a new facility depends primarily on

costs, therefore on the difference between the wage rates in a region relative to the

national average:

(2)

Labor supply in a region depends on demographic factors, and on decisions regarding

participation in the labor market. Labor supply may, therefore, be expressed as follows:

(3)

where n stands for the labor force supply in the region R over the time period t; pR,t R,t

is the activity rate in the region R over the time period t; while pop shows theR,t

movement of the population in the region R over the time period t. The

unemployment rate and the wage level affect activity rates, hence:

(4)

where u stands for the unemployment rate in the region R over the time period t.R,t

Demographic factors may be modeled in line with a migration theory, which assumes

that the population migrates from a region with a relatively low wage rate and high

unemployment to regions with a relatively high wage rate and lower unemployment,

thus:

tRtRtRtRwupop ,,2,10, ζγγγ +++=∆

1,10,, −

−=tRRtR

uw χχ

tRtRtRlnu

,,,

−=

81

(5)

The model, finally, assumes that the wage rate in a certain region over a certain period

depends on the unemployment rate in the previous period, thus:

(6)

A common assumption that the unemployment rate approximately equals the

difference between labor supply and labor demand is also incorporated, thus:

(7)

A model includes two basic mechanisms which can in time mitigate regional

differences arising from specific conditions in certain regions, or by means of which

the system overcomes such differences in the long-run. Such mechanisms are capital

mobility from one region to another and job creation in a specific region.

Data Sources

A model structured in such a way is based on a number of assumptions, most of which

have not been adequately tested empirically in the case of Croatia. First of all, the data

sources available are inadequate for a comparative analysis of fundamental indicators

usually applied in a labor market analysis at the regional level. Furthermore, most data

necessary for a thorough application of the model is lacking. While there is a general

notion that regional labor market differences in Croatia do exist, they are not

empirically verified. In addition, without empirical quantification it cannot be detected

whether the regional differences change in time or whether there is a common trend.

If differences do exist, it is necessary to investigate whether such differences will be

diminished by applying economic policy measures or whether the very structural

characteristics of the Croatian economy will enable convergence to the national

average.

The first step in an empirical analysis of the labor market is to establish whether

regional developments are correlated to developments at the national level. The paper

will attempt to provide an answer to this question by focusing solely on the

unemployment issue. The essential reason why other relevant labor market indicators

have not been addressed is the absence of appropriate data at the regional level.

The Nomenclature of Territorial Units for Statistics (NUTS). It needs to be noted that at the time this paper was5

prepared, a proposal for the division of Croatia into five regions for statistical purposes had not yet been officiallyadopted.

82

Research methods applied in distinguishing between national movements and specific

regional ones are different, thus often providing different results. Therefore, this makes

the interpretation of the data obtained in such a way even more difficult. The empirical

analysis applied in this work has followed the approach used by Shepherd and Dixon

(2002). They proved that the method of seemingly unrelated regressions (SUR) results

in more reliable estimates than the traditional OLS method.

The process of region defining in Croatia is not completed. While some regions can

intuitively be marked off according to economic or other criteria, it is clear that this is

a multi-criterion division which necessarily implies a number of solutions. At the same

time the current territorial organization, division into 21 counties, is inadequate for the

requirements of economic analysis for two main reasons:

! The first reason relates to the fact that the current territorial organization has

been in force for a relatively short time and that relating data to the previous

territorial organization is hardly possible. As the analysis can be carried out for

a relatively short time-period, a question arises as to the number of degrees of

freedom to be assigned to regression.

! Another reason relates to the fact that the objective of the results obtained by

regional analysis is often the creation of economic policy measures aimed at

resolving problems occurring in a given area. However, considering the

number of counties in Croatia, imposing specific economic policy measures

and then monitoring their effects would require considerable efforts in

developing specific criteria and gathering the relevant data in such a large

number of units.

It follows that existing territorial units need to be aggregated in appropriate regions. As

it is not the objective of this analysis to define meaningful economic regions in

Croatia, a pragmatic solution is to apply the proposal developed by a task group of the

Central Bureau of Statistics for the purpose of introducing a system of NUTS. The5

introduction of this system complies with Eurostat guidelines, which divides the EU

area according to the same principle into 78 NUTS level 1 regions and 211 NUTS level

Eurostat (1999).6

83

2 regions. As according to the NUTS nomenclature, the entire territory of the Republic6

of Croatia is just one NUTS level 1 region, for the purpose of analysis the NUTS level

2 has been applied. Subsequently, the Republic of Croatia has been subdivided into 5

regions:

! Northern Croatia (NC), which covers the following counties:

Krapinsko-Zagorska, Varaždinska, Koprivnièko-Križevaèka and Meðimurska;

! Central Croatia (CC), which covers the following counties: Zagrebaèka,

Sisaèko-Moslavaèka, Karlovaèka, Bjelovarsko-Bilogorska and the City of Zagreb;

! Eastern Croatia (EC), which covers the following counties:

Virovitièko-Podravska, Požeško-Slavonska, Brodsko-Posavska, Osjeèko-Baranjska

and Vukovarsko-Srijemska;

! Western Croatia (WC), which covers the following counties:

Primorsko-Goranska, Lièko-Senjska and Istarska;

! Southern Croatia (SC), which covers the following counties: Zadarska,

Šibensko-Kninska, Splitsko-Dalmatinska and Dubrovaèko-Neretvanska.

For the regions thus defined, the unemployment rates have been calculated according

to available data sources. The unemployment rate has been calculated as the ratio of the

number of unemployed according to the Croatian Employment Service data to the

total sum of both the employed and unemployed. The source for the number of

employed by counties is the Central Bureau of Statistics, and the data includes the total

employed by legal entities, trades and crafts, and free professionals. This data is

available only at the annual level, and presents another limitation to a potential

analysis at the regional level. In view of the above territorial division, Figure 1 clearly

shows differences between unemployment rates in individual regions. However, one

might also presume that the path of unemployment is influenced by countrywide

factors, while the level of unemployment is determined by the regional factors.

in %

0

5

10

15

20

25

30

35

40

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

NC CC EC WC SC

Figure 1 THE UNEMPLOYMENT RATE IN CROATIAN REGIONS

According to the organizational principle, when entrepreneur's headquarter is located in one region all his/her7

employees will be statistically recorded as working in this region.

84

Notes: NC - Northern Croatia, CC - Central Croatia, EC - Eastern Croatia, WC - Western Croatia, SC - Southern Croatia.Sources: Central Bureau of Statistics and Croatian Employment Service.

Since employment data is available according to the organizational principle rather

than the core activity principle, its regional distribution does not match the actual

regional distribution of economic activities. Therefore the unemployment rates7

presented in Figure 1 should be taken only as indicative and not as exact data.

Nonetheless, in order to enable the application of a selected model, regression analysis

in the remaining part of the paper uses the regional unemployment level instead of the

regional unemployment rates data. An indicator of differences in regional

unemployment rates, as a relative measure of unemployment, is much more suitable

for analysis than the unemployment level in individual regions. In order to test

whether it is possible to use unemployment levels instead of unemployment rates, a

correlation analysis for unemployment rates and levels in a region has been conducted.

The results indicate that a correlation coefficient for Northern Croatia is close to 1; for

Central Croatia it is 0.98; for Eastern Croatia 0.97; for Western Croatia 0.98 and for

Southern Croatia 0.96. All correlation coefficients are significant at the level of 5

percent. Unemployment rates and unemployment levels in individual regions are

presented in Figure 2.

Figure 2 UNEMPLOYMENT LEVELS AND RATES IN CROATIAN REGIONS

85

Source: Central Bureau of Statistics and Croatian Employment Service.

Erjavec, Cota and Bahovec (1999) report the presence of a unit root in the national unemployment series for the8

period 1992:1 to 1998:12.

Botriæ (2003) presents the results of the Phillips-Perron test, which confirm the results obtained in this paper.9

86

The application of a model by applying unemployment levels instead of

unemployment rates is not commonly used in the literature. The unemployment rate

is affected by both economic and demographic factors, such as changes in the number

of inhabitants or migrations. While the analyzed period is relatively short for recording

significant demographic changes, significant migrations due to war were witnessed in

Croatia during the analyzed period. It might reasonably be expected that such

significant migrations would affect the correlation between the rate and the level of

unemployment. In spite of the migrations, the analyzed period was short enough to

enable the switch between rates and level data. Analyses for longer periods would

require the data on unemployment rates.

The analysis was carried out on the basis of monthly data within the January 1990 -

December 2002 timeframe. Although there might be some interpretation problems

with seasonally adjusted data regressions, there are significant differences between

seasonal movements in the regional labor market in Croatia. Therefore, it seemed

justifiable to use the seasonally adjusted data for the analysis purposes. Regional and

overall unemployment level data series were seasonally adjusted separately, prior to the

regression analysis.

Main Results

The general idea was to estimate the part of the model that deals with unemployment

behavior. The first step towards this goal was to investigate the behavior of regional

unemployment data series in Croatia. Before the regression was specified, all of the

series were tested for the presence of the unit root. One of the reasons for this is that

previous studies indicated that economic data series often prove to be non-stationary.

If non-stationary series are used in regression analysis, results might be biased. In

addition, statistical tests used to assess the applied model may in such cases be less

powerful. It is worth noting that while the presence of a unit root was previously

established for the national unemployment level data, regional unemployment data has

not been examined in Croatia to date. Table 1 presents the results of the Dickey-Fuller8

test on the presence of the unit root.9

87

88

The results presented in Table 1 indicate that all the data series follow the I(1) process.

This means that the transformation of the original regional data series by using first

differences in the model is sufficient to obtain stationary series.

The next step in the analysis was to examine the series for cointegration since

non-stationary series are often cointegrated. The Engle-Granger procedure was used to

test cointegration. The results are presented in Table 2.

Table 2 RESIDUAL COINTEGRATION TESTS FOR REGIONAL UNEMPLOYMENT DATA

Dependent variable

Independent variable

CT NC CC EC WC

CT -

NC -2.88 -

CC -2.04 -2.24 -

IEC -1.69 -1.94 -1.83 -

WC -0.48 -0.94 -1.31 -1.49 -

SC -2.69 -2.84 -2.75 -2.79 0.19

Notes: Critical values: for 1% significance -3.73, for 5% significance -3.17.NC - Northern Croatia, CC - Central Croatia, EC - Eastern Croatia, WC - Western Croatia, SC - Southern Croatia, CT- Croatia total.Source: Author's calculations.

As the results indicate that there is no cointegration between regional unemployment

data and the series follows the I(1) process, the first differences of the unemployment

series were used in regression.

Before presenting the regression results it is worth noting that the applied SUR method

has its shortcomings. Although this method according to the Shepherd and Dixon

(2002) diminishes the result bias, part of the problem remains nevertheless and needs

to be taken into account in the interpretation of the regression results. This bias in the

OLS regressions, which are most often applied in the empirical research of regional

unemployment data, originates from a higher degree of correlation between regional

and national movements than that present when the SUR method is applied.

Therefore, the results obtained by applying the SUR method are more reliable. The

most evident reason for the bias to appear is the unequal size of regions. Specifically,

it is to be expected that regions which participate substantially in the overall economy

show a strong correlation with national movements. The solution to this problem is

Botriæ (2003) also reports the results of other regression equations traditionally used in regional unemployment10

data analyses. Specifically, the OLS regressions with regional dependent and national independent variables, theOLS regressions with a dependent regional variable while independent variable is used for the rest of the country,and the method of instrumental variables (IV).

89

to organize the regions in such a way that they are equally represented in the structure

of the overall economy. Unfortunately, this criterion is not applied when defining

regions in other countries either, hence this problem persists in regional analyses.

The results obtained by using the SUR method are presented in Table 3.10

Table 3 THE SUR ESTIMATES RESULTS

Dependent variables R DWIndependent variables

2

)NO )CC )EC )WC )SC

)NC - 0.46 1.770.18* 0.10* 0.29* -0.01

(5.08) (2.81) (3.77) (-0.13)

)CC - 0.60 1.380.72* 0.41* 0.80* 0.25*

(5.08) (6.25) (5.51) (3.04)

)EC - 0.36 1.730.46* 0.48* -0.52* 0.31*

(2.81) (6.25) (-2.93) (3.29)

)WC - 0.50 1.550.27* 0.19* -0.10* 0.23*

(3.77) (5.51) (-2.93) (5.80)

)SC - 0.45 1.89-0.02 0.20* 0.21* 0.79*

(-0.13) (3.04) (3.29) (5.80)

Notes: Coefficients marked * are significant at the 5% level, t-values are presented in brackets below regression coefficients.NC - Northern Croatia, CC - Central Croatia, EC - Eastern Croatia, WC - Western Croatia, SC - Southern Croatia.Source: Author's calculations.

According to the results, regions in Croatia may be divided into three groups:

1. A region where the correlation between labor market variations at the national

level and at the regional level is significant - Central Croatia.

2. Regions where the correlation between variations at the national and the

regional level is significant, but regional influences pertain - Western, Southern

and Northern Croatia.

3. Regions with strongly visible regional specifics - Eastern Croatia.

In view of the above-mentioned limitations, the fact that Central Croatia shows the

highest degree of correlation with national variations is not surprising. However,

whether there are other arguments to support these results remains to be tested. The

Gacs and Huber (2003).11

90

characteristics of the Croatian economy, in general, substantiate the results obtained,

specifically by the following arguments:

1. The structure of the Croatian economy substantially differs between the regions

thus defined. Agricultural production is most important in Eastern Croatia,

the region with the most pronounced regional specifics. At the same time,

regions characterized by traditional industrial production have a higher

correlation with variations at the national level. Such results can be observed in

other countries too. Thus Blanchard and Katz (1992), applying somewhat

different estimation methodology on annual employment data, find that the

U.S. regions with a high share of agricultural production in the 1948-1990

period exhibit a relatively high degree of independent movements, whilst

regions dominated by traditional industry were much more correlated with

movements at the national level.

2. The regional structure of the Croatian economy is also connected with a

different regional propensity to officially register one's unemployment status.

Since the transition economies on average have a higher share of the unofficial

sector in comparison with the market economies, regional distribution of the

unofficial sector could unable the comparison between the countries. The

impact of the unofficial sector on regional indicators is also evident in the

studies in other transition countries, particularly the studies addressing the

early stages of transition. However, the extent of such an impact is not easy11

to quantify. In general, there are economic activities in which unregistered

employment is relatively common. Such activities in Croatia have different

importance in the economic structure of individual regions. For example,

tourism is most developed in Western and Southern Croatia, and exhibits a

strong seasonal demand for labor; however this is not necessarily recorded in

statistical data on employment or unemployment.

3. The war in Croatia has affected the ability to implement restructuring in

certain regions. In the regions with substantial war damage economic activity

was completely suspended for a prolonged period of time. As a consequence,

labor market processes and company restructuring were postponed. That is why

Eastern Croatian unemployment is so poorly correlated with the average

Croatian unemployment movements. The war had another, more direct impact

on the data series used in the analysis. Registration presents a specific problem

More detailed results are available upon request.12

91

in the unemployment series, and so does de-registration of the unemployed war

veterans from the Croatian Employment Service records, after a while. Since

their number is not equally distributed among the regions, the figures on the

number of the unemployed in the records also impact the results.

While the analysis results obtained for Croatia are similar to those obtained by

Shepherd and Dixon (2002) for Australia, such a type of analysis does not lend itself

to a direct comparison of results with other countries. The applied method only

enables us to establish whether there is, or there is not, a difference in the degree of

correlation between the results obtained at the national level and those in individual

regions. However, by applying the Blanchard-Katz model for the EU candidate

countries Gacs and Huber (2003) tried to establish whether the movements in the

regions of candidate countries are more divergent relative to a national average than it

is the case in the EU member states. The data on five member states, notably the

Netherlands, Germany, Spain, Portugal and Italy, were used as the reference countries.

Their analysis distinguishes two groups of candidate countries - the first-wave entrants

(the Czech Republic, Hungary and Poland) and the second-wave entrants (Romania

and Bulgaria). Gacs and Huber applied the VAR model not only to the unemployment

rate but also to the activity rate, the employment growth rate and the rate of wage

growth. The results indicate that in the case of the unemployment rate, around 70

percent of the three-year forecast error results from innovations in the national

unemployment developments, leaving only 30 percent of the forecast error due to

region-specific innovations. The comparable figure for reference EU countries is 40

percent due to national factors, and 60 percent due to regional specifics. Since

according to their methodology, the regions of all the sampled countries are, inter alia,

divided into agricultural, industrial and urban, it is interesting to note that national

factors are most noticeable in industrial regions. Another rather unexpected result

indicates that urban and agricultural regions correlate equally with the national level

movements, however considerably less than the regions where industry is a dominant

activity.

The same method was applied to the available data on regional unemployment in

Croatia. Variance decomposition indicates that Croatian regions on average indicate

stronger regional influences than is the case in other candidate countries, and range

from 40 to 70 percent. This may in part be explained by the relatively larger12

Layard, Nickell and Jackman (1991, p. 471).13

92

geographical diversity of Croatia, which produces larger diversity of economic structure

among regions. Nonetheless, the characteristics of the VAR model applied, as well as

of the data available, do not allow conclusions regarding an actual ranging of

individual regions to be reached solely on the basis of such a specific model.

Conclusions

This paper presents empirical testing of the existing differences in the regional labor

markets in Croatia. The results achieved are in line with our expectations, indicating

that substantial differences between Croatian regions regarding unemployment do

exist. Three groups of regions have been identified. Central Croatia shows the highest

degree of correlation with the movements at the national level. As for the comparison

with other countries, it has been established that regional impacts on the labor market

in Croatia are much more pronounced than in some EU candidate countries.

As it is among the objectives of regional analyses to establish a need to introduce

economic policy measures at the regional level, it needs to be noted that the analysis

results obtained in this paper may not be used for this purpose for a number of

reasons. One of the most important ones is that the regional structure presented here

is no more than a proposal for a possible division that was drafted solely for statistical

purposes, in other words for data collection and processing. Although the introduction

of economic measures requires prior analytical estimates, and quality data sources, it

is not necessary for the regions to be aggregated in the same manner for statistical and

for regional policy purposes. Furthermore, despite public pressure to resolve specific

local problems in the labor market, the literature does not advocate the necessity of

imposing specific regional policy measures in the labor markets. Specifically, the13

unemployment rate is not always determined by a process of optimal resource

allocation, but is very often influenced by the unemployment insurance system and the

system of wage setting. Both systems have the potential to generate adverse externalities

and may thus lead to the unemployment rates above an economically optimal level. A

regional approach to the resolution of labor market problems, without adequate

analytical background, may intensify adverse externalities. At the same time, the focus

on a regional approach and regional research persists in all EU candidate countries,

93

inter alia, because of access to EU structural funds. Since one of the EU goals is to

achieve higher employment rates, it may be expected that labor market related regional

analyses will in future be of particular interest. At present it is hard to offer any

concrete recommendations as to whether and in what way to intervene by special

measures at a regional level.

For regional labor market developments analysis to be improved, in addition to

considering the differences in unemployment, other parameters such as employment

variations, differences in wages and activity rates, and a number of others must be

taken into accounts. A particular issue that may affect the speed of balancing

unemployment rates between Croatian regions is the mobility of capital and labor

force. In Croatia there is a tendency towards higher concentration of economic

activities in some parts of the country; while labor force mobility remains relatively

low. These events are certainly not related only to the economic factors. As the

analytical apparatus does exist and has been presented in this paper, limited data

sources are responsible for not being applied more thoroughly. Once other labor

market segments are analyzed, it will be possible to understand regional labor markets

in Croatia better and determine whether some action at the regional level needs to be

taken.

94

References

Amendola, Adalgiso, Floro Ernesto Caroleo and Gianluigi Coppola, 2003, "Regional

Differences in the European Labour market", a paper presented at the "Peripheries, Centres

and Spatial Development in the New Europe" conference organized by the European

Regional Science Association, Jyväskylä, Finland, August 27-30, 2003.

Bertola, Giuseppe, Tito Boeri, and Giuseppe Nicoletti (ed.), 2001, Welfare and Employment

in a United Europe, Cambridge, Massachusetts: The MIT Press and Fondazione Rodolfo

Debenedetti.

Blanchard, Olivier Jean and Lawrence F. Katz, 1992, "Regional Evolutions", Brookings Papers

on Economic Activity, 1, pp. 1-61.

Botriæ, Valerija, 2003, "Regional Differences in Unemployment: the Case of Croatia", a paper

presented at the "Peripheries, Centres and Spatial Development in the New Europe"

conference organized by the European Regional Science Association, Jyväskylä, Finland,

August 27-30, 2003.

Central Bureau of Statistics of the Republic of Croatia, Statistièka izvješæa - Zaposlenost i

plaæe (Statistical Reports - Employment and Wages), various issues, Zagreb: Central Bureau

of Statistics of the Republic of Croatia.

Engle, Robert F. and W. J. Granger, 1987, "Co-integration and error correction:

representation, estimation and testing", Econometrica, 55, pp. 251-276.

Engle, Robert F. and Byung Sam Yoo, 1987, "Forecasting and Testing in Co-integration

Systems", Journal of Econometrics, 35, pp. 143-159.

Erjavec, Nataša, Boris Cota and Vlasta Bahovec (1999): "Monetarno-kreditna i realna

privredna aktivnost u Republici Hrvatskoj: VAR model" (Monetary-Credit and Real Economic

Activity in the Republic of Croatia: VAR Model), Ekonomski pregled, 50 (11), pp.

1488-1504.

Eurostat, 1999, "The Nomenclature of Statistical Territorial Units - NUTS"

http://europa.eu.int/comm/eurostat/Public/datashop/print-catalogue

95

Gacs, Vera and Peter Huber, 2003, "Quantity Adjustment in Candidate Countries Regional

Labour Markets", a paper presented at the "Peripheries, Centres and Spatial Development

in the New Europe" conference organized by the European Regional Science Association,

Jyväskylä, Finland, August 27-30, 2003.

Greene, William H., 2000, Econometric Analysis, 4 edition, London: Prentice-Hallth

International.

Croatian Employment Service, "Nezaposlene osobe po županijama", (Unemployed Persons

by Counties), http://www.hzz.hr/docslike/statistike/tablica%2027.xls.

Layard, Richard, Stephen Nickell and Richard Jackman, 1991, Unemployment:

Macroeconomic Performance and the Labour Market, Oxford: Oxford University Press.

Shepherd, David and Robert Dixon, 2002, "The Relationship between Regional and National

Unemployment", Regional Studies, 36(5), pp. 469-480.